Clinical Social Work Research

Hi All
Surprise, surprise! 1 more lecture and we are half way through the course. Let's review today's lecture.

(1) We reviewed concepts of association and causality .

(2) We talked about what a PRE measure was about.

(3) We reviewed SPSS cross-tabs and how to get the different association measures in SPSS.

(4) We defined the general concept of a PRE measure and informed vs. uninformed guessing.

(5) We defined the PRE measure lambda and discussed how to calculate it and interpret it. We also defined E1 and E2 errors used in the calculation of lambda and how to calculate them. We observed that lambda is good when you have nominal data.

(6) We defined the PRE measure gamma and discussed how to calculate and interpret it. We observed that it required us to understand the concept of concordant pairs and discordant pairs and how to calculate them. We observed that gamma is good when you have ordinal data.

(7) We used SPSS to introduce the concept of a scatter graph to begin to understand relationships.

Next week please read the chapter on regression. Continue to frame your hypotheses in terms of contingency tables and the questions you intend to ask.

Have a good one
Tarynn

Posted by tmwitten at 09:22 AM | Permalink | Comments (0)

Hi All
Hope you are having a good Turkey Day and are continuing to review the material you have not had a chance to really digest (along with the turkey). Remember to be formulating questions, based upon your hypotheses, that can be tabled like contingency tables. See you all in a week.

Tarynn

Posted by tmwitten at 09:12 AM | Permalink | Comments (0)

Hi Everyone-
Wow! What a day that was. We covered a whole bunch of new concepts but we really didn't do much math. Whew! So let's see what we are supposed to have taken home from this week's lecture.

(1) We reviewed construction of a contingency/cross-tab table and what all of the different parts of the table mean. We also covered how to interpret the different components.

(2) We discussed independent/dependent variables and how to read the question a table was asking.

(3) We covered how to compute row, column, and total percent and how to interpret what the different cells and combination of cells mean under the different percentage calculations.

(4) We saw a way to illustrate a contingency table graphically using mosaic displays .

(5) We covered the concepts of odds, conditional odds, and odds ratio , what each means and how to interpret each of them.

(6) We reviewed how to make SPSS do contingency table/cross tab analysis. And we discussed how to interpret the output results.

(7) We then looked at how to determine the existence of an association between variables and we introduced idea of a the bubble plot as an easy way to begin to interpret contingency tables and to look for associations.

(8) We followed this with a discussion on different contingency tables and bubble plots and what they would look like for different types of association. We illustrated positive and negative association and how we could use the odds ratio to help us see if an association exists. And we finished this by discussion the idea of strength of a relation and what type of scale is usually used. We also talked about linear and nonlinear associations.

(9) We closed our discussion with the concept of causality . We discussed what causality meant and what some of the requirements were to have a causal association, and we introduced the idea of a spurious relationship. We also illustrated how to use SPSS layers in Cross-Tab to help determine if a spurious relationship exists.

Don't forget to reread your chapters during the week that I will not be here. When I return, we will be working hard to finish the material on measures of association and move on to linear regression. Read in the textbook about the association measures lambda and gamma because we will be covering them in detail. Those of you that did not hand in your Step 2 paper, make sure you have it ready for the next class.

Please begin to think about how contingency tables might apply to your research effort. Think about your measurement instruments and other information that you will be recording and how you might formulate questions in terms of contingency tables as they relate to your hypotheses. Also, begin to seriously consider the question of how you will create your SPSS database.

Have a great next two weeks
Tarynn

Posted by tmwitten at 08:46 AM | Permalink | Comments (0)

Hi Everyone -
My apologies for the delay in posting this week's commentary. I have been out of town and did not have easy internet access. So let's review what we have learned about this week.

In Week 11 class

(1) We covered a detailed example of how to go through analyzing the distribution properties of a specified variable. We covered all of the measures of central tendency and we covered all of the measures of dispersion. We looked at how different forms of subgrouping can generate different outliers and we discussed the impact of this on how we analyze our data. We used SPSS to go through the complete analysis. For those of you that missed this, you should consult with a friend in the class.

(2) We briefly reviewed the homework problem solutions for the dispersion homework.

(3) We began a discussion of cross-tabulation or contingency tables and used SPSS to construct the tables. We discussed why one would use cross-tabs/contingency tables and what the basic structure of a contingency table looked like.

(4) We defined the basic components of an r x c contingency table were: cell, cell frequency, cell percent, row and column marginals and total.

(5) We discussed how to interpret contingency tables based upon how they were constructed .

(6) We covered how to use the contingency table mode in SPSS to construct contingency tables and we examined the output of some simple examples.

(7) We discussed issues of thinking about our database construction as we move towards our IRB proposal completion and how the types of questions that we wished to pose would affect the way we set up our database examples. We discussed how string variables and coding variables could be used to help us to keep track of subcases if we wished to analyze our data using different subcases.

This coming week we will continue our discussion of contingency tables, measures of association and how to use SPSS to analyze the results of a contingency table analysis.

Have a great one
Tarynn

Posted by tmwitten at 12:26 PM | Permalink | Comments (0)

Hi Everyone-

Do you realize that we are now 2/3 of the way through the first semester and 1/3 of the way through the academic year! And wow! What an intense and dense lecture that was. Lots of material to learn. And we sure covered a wide variety of new equations and concepts. So let me see if I can summarize what you should have taken home from this lecture.

In this class we covered the detailed concept of dispersion or spread of a distribution.

(1) We reviewed the range and discussed why it is a very rough descriptor of the dispersion of a distribution.

(2) We introduced the concept of the IQV - Index of Qualitative Variation as an alternative descriptor of variation that is particularly useful when one wishes to compare between groups whose categories are nominal descriptors.

(3) We next introduced the concept of variance and standard deviation . We discussed the idea of deviation from the mean value and how summing the squares of the deviations of each value from the mean would be a useful measure of the total variation of the distribution. This gave us the concept of the sum squared deviation or SS . We then used this to demonstrate how to compute the variance and standard deviations. We talked about how there were different notations for when one is computing the variance of the total population versus a sample of the population. And we also covered some alternative formulae for computing the mean and variance because there are problems using the definitional equations.

(4) We then discussed the meaning of the variance/standard deviation in terms of area under the "bell curve" and how that is useful. We examined, using the bell curve, what happens when the standard deviation is small versus when it is large.

(5) We then looked at how to compute the standard deviation and the variance from grouped data . We observed that the formula for doing this is methodologically similar to the way we computed the mean for grouped data.

(6) Next we introduced the concept of the IQR - Interquartile Range . We defined it and discussed why it is a more stable range value than the actual range. First we considered how to compute the IQR using the exact data. We then considered how to construct the IQR by revisiting the formula for constructing the median quartile using grouped data. We did some examples where we computed Q25 and Q75 using the grouped data and we demonstrated how the formula for Q50 (median) could easily be altered to compute the "tile" data for any desired "tile".

(7) We next covered boxplots and modified boxplots . We discussed how they are constructed and how to interpret them.

(8) We addressed the question of how one ethically decides when a data point or points can be considered to be an outlier. To do this, we defined the constructs of upper fence and lower fence . We then stated the rule that a datapoint that falls outside the fence may be considered to be an outlier once you have checked to make sure that you have entered the data correctly from the data sheets and that there is no problem with the data in the database.

(9) We then discussed the problem of scales. We observed that some data can be on a different scale of magnitude than another. We observed this when we did clustered bar graphs and histograms and discovered that the frequency scales were quite different in our example. We talked about how one can "lie" with graphs and how to avoid this problem by scaling to percentages. We also introduced the concept of CV - Coefficient of Variation as a way of scaling the variation so that distributions that are at different scales can be compared.

(10) We closed our discussion with a Halloween scare from Tarynn who showed us the actual equations for calculating skewness and kurtosis . But we were reassured that we would not actually have to use them. We discussed how to interpret the skewness and kurtosis measures and what the standard error of skewness and kurtosis meant.

(11) In SPSS we went over some new features of the bar graph and we discussed how to create a boxplot . We also addressed how to enter data into an SPSS spreadsheet in a way that makes it more amenable for plotting boxplots. We also went over the remainder of the measures of dispersion in the Analysis component of the SPSS modules we have previously covered.

Don't forget to do your practice problems. Have a great week.
Tarynn

Posted by tmwitten at 08:43 AM | Permalink | Comments (0)