Final analysis & Discussion_Volen
- 1. JANET VOLEN A RESEARCH PROTOCOL Prepared by: Dr. Alfonso Scandrett Jr Medical Epidemiologist 6317 High View Road Greensboro, North Carolina 27410 Research Investigator: Ms. Janet Volen October 4, 2011 Ms. Janet Volen of Columbus, Ohio, has asked me to complete the descriptive analysis for the project entitled: “The use of the Framingham Risk Score to initiate lifestyle modification and for patient education, by identifying men and women at increased risk for future cardiovascular events.” I have provided the following objective analysis: I have undertaken the preliminary research and will provide recommendations to the client. Fully addressed and provided the client with the scope of the project Individual consultation on research project Provide suggestions for possible improvements to your methodology and instrument. The appropriate statistical procedures were used based on the nature of the project and the hypothetical outcomes. I have actively input data on to a tally sheet for stat evaluation using the “MYSTAT” statistical analysis program. I have run statistical analysis of your data upon your request. I have helped with the statistical interpretation of that data upon your request. (As it is not in my scope of expertise, I was not instructed to write or assist in the writing of the final thesis or project paper for this agreement).
- 2. Materials 1. I have sent an analysis that is descriptive in nature. 2. I have sent under a separate email cover graphs and statistical data runs (optional to keep) 3. I have sent under an email mailing a “EXCEL” copy with attachments that should enhance this descriptive analysis and should be included as graphs and charts in your final presentation. 4. Knowing that you need to write the final paper I will leave up to you how you feeland would like the supportive materials to be used. Of course, you can also develop your own as well. ANALYSIS OF THE PROJECT Goal: 1. The Framingham Score will be used as the dependent variable to conduct lifestyle changes for patient education by identifying men and women who are at risk for future cardiovascular events. Methodology Along with the descriptive information of all research categories,a reliability coefficient was performed. Standard statistical measures (means,standard deviations) are reported. To determine possible interactions and main effects between demographic and clinical variables, a two-way analysis of variance and student T-Test was completed. Other analysis such as regression analysis, factorial analysis and discriminate analysis were also reviewed for possible use. The evaluation process was begun with the examination of two data groups: demographics and clinical. The demographic variables were: Age, Gender, Race,Marital Status, Income, Education, Employment and Living conditions as describe in Table 1. The clinical variables covered Age, Total Cholesterol, HDL,Systolic Blood Pressure,the Framingham and Comparable Risk scores. The mean scores,standard deviation with confidence intervals, and range were also provided in Table 1. P-Values were calculated at the .05 significant levels for each variable. A significant P-value suggests a degree of heterogeneity among those variable whereas those variable that were not significant suggest a more homogeneity in nature in the responses of those subjects. However,a degree of caution regarding generalizing a statement about these variable should be considered because of the small population number (n=50). Demographics: The total number of participants (n=50) were measured over a range of demographic variables (gender, age, sex, ethnicity, marital status,education, income, and Living situations). Specifics to the different types of response can be viewed by examining the demographic sheet. A number of variables showed significant p-values. Which suggest an overall mean score difference among the subjects of that particular variable and their response? The Story Line: To determine if any relationship existed between research categories,a correlation matrix was designed. Either the variable strength or the role they play directly or indirectly in to the research question is important. To determine strength a Pearson Moment Correlations is used. The cut off for practical use this study was set at r=.30 which is based on Witte’s (1993) recommendation for making correlation comparisons (See CORRELATION MATRIX). However values that were >r=.40 were only reported. I am reporting that the relationship between research categories was low although statistical significance was obtained in a number of the relationships. My overall goal was to identify patterns among the variables. Please note the following relationships:
- 3. PCS 2 – PCS1 .91 HDL – PCS2 .43 HDL – PCS1 .46 AGE – ComRisk .60 MCS2 – MCS1 .92 TChol – SYS_BP .46 FramScore – Tchol. .63 HDL – FramScore .56 FramScore-ComRisk .56 NOTE:It is very interesting to me that your intervention, which is the Framingham scores,did not correlation well with both your pre and posttest components. However,to save face I will say that the .3 correlations that I failed to list with PCS1 and PCS2 are fine. I am just mentioning to you that that later in my discussion I point out some issue regarding this especially in the area of internal consistency and t-test. Make a note of this observation. A demographic analysis using all variables was conducted to be able to determine size and strength of each variable and the role that each variable might play in its relationship to the research question. (See “SOCIODEMOGRAPHIC ANDCLINICALVARIBLES”) Bar graphs are provided (see my email) to show visual demographics of selected demographic variables and clinical variables. Equally Chi-Sq cross tabs were used in order to demonstrate variable associations. The P-value that is seen with the Chi-Sq was set at the .05 degree of significances. Variables were treated as nominal and categorical class variables. Coding on the excel data sheet (see data sheet) shows the coding that was used in order to describe each variable category. The pre and post variables of MCS1, MCS2, PCS1 and PCS2 were set in a single factor (one-way) ANOVA to determine the variance between them. (See TABLE A). Table A demonstrates a non- significant relationship between these variables. This means that the means of each variable is close in nature and do not suggest any special conditions that would highlight any one particular variable as being a major player in your study. In nonprofessionals’ terms, this would mean that your posttest response is not a factor after your intervention. A factorial analysis was conducted to determine those variables that would be good indicators for proving the research hypothesis. Both demographic and clinical variables were loaded into the factorial analysis. The process produced seven factors; two factors presented Eigen values of 1.4 to 2.9. The other five factors were too weak to mention. Its Eigen Value indicates the strength of a factor. Eigen Values of less than one are usually not interpreted. It is important to mention that all variables except for MCS2 loaded very strongly onto the first factor that had an Eigen Value of 2.9l explaining 41.5 percent of the variance. The p-value to prove that there was one or more variables that did not load was p=.000 with 21 degrees of freedom. The second Eigen Value equaled 1.4 which itself explained 20.3 percent of the variance. So what does this all mean? The number of variables that load on that particular factor generates a factor’s strength. For explanations sake:It is kind of like a mathematical vote of support. Variables are loaded in order of their mathematical perceived strength. You have two that show their strength in regards to the dependent variable. Both of these Eigen values explain 61.8 percent of the explained variance, which means that we can attest to why certain things are happening regarding you, hypothesis but we also know that there is 49.2 percent of the variance that is not explained. So if 61.8 percent is explaining occurrences over 2/3 of the time then can say that accepting the null we would maybe be able to explain what happening 2 out of 3 times. Are those odds that one would want to gamble on…..???
- 4. Those factors are then place into a Regression Analysis. In this study’s case Regression Analysis were run. Each was given a DEPENDENT variable. The dependent variable used was the variable that was suggested in the research question as ones that should be considered. Regression 1 had PCS2 as a dependent variable and Regression 2 had MCS2 as a dependent variable. The goal here is to determine which of these INDEPENDENT variables explains the most variance as it relates to answering your research question. The higher the variance means the better the chance of explaining away the variance of the alternative Hypothesis. A lower variance means the greater the chance of rejecting the original hypothesis. For explanation sake picture going to jail and having witnesses who come forth to speak positively on yourbehalf.In orderto determine if those witnessesare strong enough in your favorwe run a factorial analysisthen we put themon the witness stand to see how much they truly do know in your favor… thisis the regression. Do you like my analogy? o JANET> Please know that in the above explanation it is important for you to understand that factors in a factor analysis are to determined (operative word here is determined)which of your variables when placed together with other variables are strong enough to be used to explain why something is happening. In other words if your group had a significant average score increase due to your intervention there are a factor of variables working togetherin that intervention were the ones that caused it. HOWEVER, the factors in a regression are seeking to be more specific, by identifying those specific independent variables that to explain variances. Now with the above being said… Check out the following dialogue>> Looking at PCS2 as the dependent variable …..the regression analysis found that the following variables Comparative Risk, Framingham Scores,Systolic BP, HDL and Total Cholesterol explained approximately 26 percent of the variance of the Dependent Variable. The question is what happened to the rest of the variance. The remaining 74 percent is not explained and could suggest a Type-I error if hypothesis is accepted. The MCS2 (Mental Health) variable only explains 16 percent of the variance which leave 85 percent un-explained. To take the analysis a bit further what is being said here is that all of the variables mentioned above are combine and through a regression analysis, they do a very bad job in explaining the dependent variable of MSC2 and PCS2. The unexplained error in both of the analysis could be due to a multitude of reason that is not explained here. A type I error, also known as a false positive, occurs when a statistical test rejects a true null hypothesis (H0). For example, if a null hypothesis states a patient is healthy, and the patient is indeed healthy, but the test rejects this hypothesis, falsely suggesting that the patient is sick. The rate of the type I error is denoted by the Greek letter alpha (α) and usually equals the significance level (or size) of a test. A type II error, also known as a false negative, occurs when the test fails to reject a false null hypothesis. For example, if a null hypothesis states a patient is healthy, and the patient is in fact sick, but the test fails to reject the hypothesis, falsely suggesting that the patient is healthy. The rate of the type II error is denoted by the Greek letter beta (β) and related to the power of a test (which equals 1-β). The desired (i.e., non-erroneous) outcomes of the test are called true positive meaning "rejecting null hypothesis, when it is false" and true negative meaning "not rejecting null hypothesis, when it is true". A statistical test can either reject (prove false) or fail to reject (fail to prove false) a
- 5. null hypothesis, but never prove it true (i.e., failing to reject a null hypothesis does not prove it true). In colloquial usage type I error can be thought of as "convicting an innocent person" and type II error "letting a guilty person go free". Based on the hypothesis statement a T-Test was employed to either accept or reject the null hypotheses of the research question. However,it should be noted that the T-test is not as rigorous of an evaluations tool as the ANOVA. I need to point out that the ANOVA seems to be given more of an acceptance of the null hypothesis where the t-test is showing MENTAL Health (MCS1 and MCS2) have significant alpha. I did not run a post hoc test when looks at interactions between variables because of this. The variable used in this analysis were: Instrument Reliability Here we are seeking to determine if the survey instrument was reliable. To determine if the research categories were reliable, a Cronbach Alpha that estimates internal consistency was performed on the pre and posttest variables (MCS1, PCS1,MCS2, and PCS2). The results showed an internal reliability coefficient of .69. Whereas when just the internal components of MSC1 and MSC2 were combined, they yield a higher alpha of .94 and PCS1 and PCS2 together had an alpha of .94. From a reporting standpoint, this is excellent which means there is a strong reliability that your post-test responses based on your intervention are measuring what you want them to measure. However,this does not mean that your subjects are understanding and buying into what you are saying. Interestingly the Framingham Scores when connected to the PCS2 response their alpha was consistent (.-96) but in a negative fashion. This was not the case when the Mental Health responses were combine and examined (alpha=.01). These outcomes suggest that the items in your Framingham score may be reliable and consistent with the items in your Physical Health approach but not in the manner, you may want. This also may suggest that the items in your Framingham scores when examined with Mental Health they are not consistent. Note that one observation could be that the Framingham scores are physical in nature like those of Physical Health, whereas the Mental Health is not. Just a point to look at. T-test However,the results of the t-test do not back up my above observation. The results of the t-test suggest that your middle-aged subjects are more incline mentally then they are physically to response to your intervention. I think being a nurse you might agree that it is easier to think about being healthier rather than doing things to bring down one’s cholesterol level or BP or HDL and risk factors. Given the nature of your populations, I would say this might be the case,especially if they are sedentary. Janet what do you think??? Hypothesis #1: Ho: > There will be no measured effect of reducing the risks of cardiovascular disease among mid-life adults (men/women) when a low-intensity lifestyle intervention is introduced. Ha:> There will be a measured effect of reducing the risks of cardiovascular disease among mid- life adults (men/women) when a low-intensity lifestyle intervention is introduced.
- 6. Hypothesis: #2: Ho:> The use of the SF-36v2 health and well-being survey is not an effective means for measuring the Framingham Study as an intervention in reducing the risk of cardiovascular disease. Ha:> The use of the SF-36v2 health and well-being survey is an effective means for measuring the Framingham Study as an intervention in reducing the risk of cardiovascular disease. Discussion and Recommendation: Based on the aforementioned evidence, it is my opinion that regarding both research hypotheses 1 and 2, you should not accept the alternative hypothesis that there are significant changes in the way your subjects responded to your interventions. Although there were some significant values shown in the different analysis, the proof is clear those significant changes were at best borderline and acceptances of either hypothesis based on one significant, t-test may indicate a type-1 error. To further support, this claim I conducted a non-parametric test called Mann Whitney. This test is similar to the ANOVA that is used for parametric analysis. The response to those analysis are presented in your exceldocuments. In doing, this I believe you will find that even the Mann Whitney shows no significant differences in most of your variables except a few. Briefly when examining both males and females the results seem to suggest that there is a variance in the Framingham Intervention when gender is considered. Gender also seems to be a factor among Comparative scores. A cross-tabulation would show which questions are causing the concern if those data were available. Perhaps for another study. HDL seems to be an issue among the four African-Americans of the study. Recommendations: Although significant changes in the data relative to this project were not sufficient because of the study design: It should be noted in your discussion that the instruments used (your survey) has been validated in a previous studies where the population and parameters were stronger and could be used in future clinical research. This should go in your recommendation or restrictions chapter. It is also recommended that this study be continued but with larger populations numbers from a more diverse and randomly selected population. The manner in which the intervention was used may not have provided the means and methods needed to show a significant mean response in the post-test. The related literature does show the Framingham model as a strong and robust instrument that has been used to measure health behaviors and health outcomes. I would recommend if this research were going to be further examined that the research statement be revamped to state the problem. It present form is confusing. The individual categories that make up the Mental and Physical Health outcomes will be very helpful in future analysis as they would yield a better and specific picture as to the borderline data that is being presented.